Using ICA Method To Extract Regional Hydrological Signals Of GRACE Time-varying Gravity Field

Water resources is very important to the survival of humans,studying the change of regional land water reserves can provide a clearer understanding of the change characteristics of water storage variations in a region,which can provide better help for the study of local drought,flood and other natural disasters,and also provide basic information for the protection of sustainable development of water resources in this region.In this paper,GRACE RL06 time-varying gravity field model data published by CSR(Center for Space Research)from January 2003 to December 2014 were used to invert regional land water changes,in addition,principal component analysis and independent component analysis were used to extract the change signals of land water reserves.Combined with various data,the change characteristics of regional water reserves were comprehensively analyzed.The main work is as follows:1)Simulation experiments are designed to verify the capability of PCA decomposition and ICA decomposition methods in signal decomposition,and it turns out,both methods can decompose each signal into different components according to the strength of the analog signal,however,in addition to the main signals,components decomposed by PCA are also affected by other signals,and ICA method solves this problem well.According to the results of ICA decomposition,this method can completely separate each signal,that is,each component contains only one analog signal.Comparing the results of PCA decomposition and ICA decomposition,it can be seen that the ICA decomposition method based on fourth-order statistics is better than the PCA decomposition method based on second-order statistics in terms of signal decomposition.2)Using GRACE time-variable gravity field model data to calculate the Okavango delta in Africa and the surrounding areas of water reserves,the mass change were extracted by principal component analysis and independent component analysis respectively,it is compared with the GLDAS hydrological model.The results show that the water reserves in the northeast of Okavango river basin show periodic changes,the correlation coefficient of the time series corresponding to GRACE-IC1 and GLDAS-IC1 of two kinds of spatial feature distribution signals in the same position reaches 0.85.The water variations in the Okavango delta area increased from January 2003 to October 2011,the correlation coefficient of the time series corresponding to GRACE-IC2 and GLDAS-IC3 of two kinds of spatial feature distribution signals in the same position reaches 0.81.It indicates that the inversion results of GRACE and GLDAS hydrological model have a strong consistency in the research area.In addition,GPCC precipitation data and WGHM hydrological model data are introduced to analyze the reasons for the variation of terrestrial water reserves in the study area.Experimental results show that compared with the traditional polynomial fitting method,the independent component analysis can directly extract the spatial-temporal characteristics of the quality change signals in a specific location in a large area.By comparing the third component of the analysis results of the two GRACE methods,it can be seen that the independent component analysis has stronger decomposition ability for signals than the principal component analysis.3)Using GRACE data to calculate the change of water reserves in the three gorges reservoir area in China,according to the actual data of Chongqing Water Resources Bulletin and the rainfall data of GPCC,it can be seen that the abnormal decrease of rainfall in the three gorges reservoir area in 2006 led to the extremely severe drought in this area.The calculation results of GLDAS hydrological model data and WGHM hydrological model data show that this extremely severe drought has not only caused a sharp decline in surface soil water in the three gorges reservoir area,but also caused a certain degree of loss to the groundwater in this area.After deducting the soil water composition of GLDAS from the water reserves calculated by GRACE,the changes of water reserves caused by three representative impoundments of the three gorges reservoir were analyzed;then,ICA method was used to decompose and reconstruct the water reserve signal in the three gorges reservoir area.After deducting the soil water composition of GLDAS,the water reserve changes caused by three representative impoundment of the three gorges reservoir were analyzed again,the results showed that the average equivalent water height increased by 28 mm in the second stage compared with the first stage.Compared with the second stage,the average equivalent water height of the third stage increased by 17 mm.The water reserves in the second stage increased compared with the first stage,but showed a slow decline in the second stage,and the water reserve in the third stage is relatively stable.Combined with the actual situation,the results obtained by ICA decomposition and reconstruction are more reasonable than those obtained directly by GRACE data.